pith. sign in

arxiv: 2607.00710 · v1 · pith:Z2UUQJ27new · submitted 2026-07-01 · 💻 cs.CV · cs.AI· cs.RO

Creating Impactful Autonomous Driving Datasets: A Strategic Guide from Research Gap to Benchmark

Richard Schwarzkopf , Jonas Merkert , Frank Bieder , Annika B\"atz , Alexander Blumberg , Carlos Fernandez , Felix Hauser , Fabian Immel
show 17 more authors
Christian Kinzig Hendrik K\"onigshof Fabian Konstantinidis Martin Lauer Willi Poh Nils Rack Kevin R\"osch Yinzhe Shen Marlon Steiner Gleb Stepanov Dominik Strutz \"Omer \c{S}ahin Ta\c{s} Julian Truetsch Kaiwen Wang Royden Wagner Jan-Hendrik Pauls Christoph Stiller
This is my paper

Pith reviewed 2026-07-02 14:13 UTC · model grok-4.3

classification 💻 cs.CV cs.AIcs.RO
keywords autonomous driving datasetsdataset designresearch gap diagnosisdata operatorsKITScenesannotation strategybenchmark creationresource-efficient data collection
0
0 comments X

The pith

Impactful autonomous driving datasets begin with diagnosing whether a research question faces a data problem or an evaluation problem, then applying the cheapest operators to close the gap.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper argues that well-designed datasets advance autonomous driving research only when creators first identify the exact blockage in a research question. Once the blockage is classified as data-related or evaluation-related, the next step is to pick the least costly sequence of operations that resolves it, turning to fresh recordings solely as a last resort. This process is illustrated by tracing how major existing datasets evolved and is put into practice through the authors' KITScenes dataset family. Resource-constrained groups gain a clear decision path that avoids defaulting to expensive new data collection. The framework organizes choices across gap diagnosis, operator selection, sensor configuration, and labeling tactics.

Core claim

The central claim is that impactful dataset creation begins with a diagnosis: whether a research question is blocked by a data problem or an evaluation problem, and proceeds by selecting the minimal data operator(s) that closes the resulting gap, recording new data only when no cheaper operator(s) suffices. The authors analyze the evolution of major autonomous driving datasets through this lens and distill a strategic framework spanning gap identification, operator choice, sensor suite design, and annotation strategy, which they ground in their KITScenes case study.

What carries the argument

The diagnosis step that distinguishes data problems from evaluation problems, followed by selection of the minimal data operator(s) needed to close the identified gap.

If this is right

  • Dataset projects should begin by stating the precise research question and classifying its blockage before choosing sensors or collection methods.
  • Re-annotation, augmentation, or re-use of existing recordings should be evaluated first whenever they can close the gap at lower cost.
  • Sensor suite and annotation decisions follow from the chosen operators rather than preceding them.
  • Analysis of past datasets reveals which operator sequences produced lasting benchmarks and which did not.
  • Smaller teams can allocate resources more predictably by treating new data recording as the final rather than default option.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same diagnosis-plus-minimal-operator logic could be tested in non-driving domains such as medical imaging or natural language datasets to check transferability.
  • Impact metrics such as citation patterns or downstream algorithm improvements could be compared between datasets that followed the framework and those that did not.
  • The framework implies that many existing datasets may contain excess data whose collection could have been avoided by earlier operator choices.

Load-bearing premise

The primary and generalizable method for creating impactful datasets is this initial diagnosis of the blockage type followed by minimal operator selection.

What would settle it

A dataset created without performing the gap diagnosis or without restricting itself to minimal operators that nevertheless produces higher research impact than comparable datasets built with the process would undermine the claim.

Figures

Figures reproduced from arXiv: 2607.00710 by Alexander Blumberg, Annika B\"atz, Carlos Fernandez, Christian Kinzig, Christoph Stiller, Dominik Strutz, Fabian Immel, Fabian Konstantinidis, Felix Hauser, Frank Bieder, Gleb Stepanov, Hendrik K\"onigshof, Jan-Hendrik Pauls, Jonas Merkert, Julian Truetsch, Kaiwen Wang, Kevin R\"osch, Marlon Steiner, Martin Lauer, Nils Rack, \"Omer \c{S}ahin Ta\c{s}, Richard Schwarzkopf, Royden Wagner, Willi Poh, Yinzhe Shen.

Figure 1
Figure 1. Figure 1: Publicly available autonomous driving dataset with sensor data and labels [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Historical SOTA progression of representative online HD map construction [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: KITScenes Multimodal [1] sensor setup. Our sensor rack (left) is depicted along [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
read the original abstract

Well-designed autonomous driving datasets have fundamentally shaped research progress, yet existing literature primarily describes what datasets contain rather than how to strategically design impactful ones. This is especially limiting for small and medium-sized labs and startups that cannot afford to misallocate scarce resources. We argue that impactful dataset creation begins with a diagnosis: whether a research question is blocked by a data problem or an evaluation problem, and proceeds by selecting the minimal data operator(s) that closes the resulting gap, recording new data only when no cheaper operator(s) suffices. We analyze the evolution of major autonomous driving (AD) datasets through this lens and distill a strategic framework spanning gap identification, operator choice, sensor suite design, and annotation strategy. We ground the framework in a running case study of our KITScenes dataset family. The datasets are available at: https://kitscenes.com/

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper claims that impactful autonomous driving (AD) dataset creation begins with diagnosing whether a research question faces a data problem or an evaluation problem, followed by selecting the minimal data operator(s) to close the gap, with new data recording used only when cheaper operators are insufficient. It supports this by retrospectively analyzing the evolution of major AD datasets through this lens, distilling a framework covering gap identification, operator choice, sensor suite design, and annotation strategy, and grounding the approach in a case study of the authors' KITScenes dataset family.

Significance. If the proposed diagnosis-plus-minimal-operator framework holds and generalizes, it could help smaller labs and startups allocate resources more efficiently when creating AD datasets, potentially increasing the rate of targeted, high-impact contributions. The paper's open release of the KITScenes datasets is a concrete positive contribution that enables community follow-up.

major comments (3)
  1. [§4 and §5] §4 (framework distillation) and §5 (KITScenes case study): The central claim that the diagnosis step plus minimal-operator selection reliably yields more impactful datasets than alternatives rests on post-hoc reframing of existing datasets and a single self-authored case study. No forward test, controlled comparison against alternative design processes, or external replication is reported, so the optimality of the minimality criterion remains unverified.
  2. [§3] §3 (evolution analysis): The mapping of historical dataset decisions onto the proposed 'data operator' taxonomy is presented as evidence for the framework, but the taxonomy itself is introduced in the same section; this creates a risk that the analysis is shaped by the framework rather than independently motivating it.
  3. [§2] §2 (gap identification): The distinction between 'data problem' and 'evaluation problem' is introduced without an operational, reproducible procedure or decision criteria; without such a procedure the diagnosis step cannot be applied consistently by other teams, undermining the claim that the framework is strategic and generalizable.
minor comments (2)
  1. The abstract states that the datasets are available at https://kitscenes.com/; the manuscript should include a permanent DOI or archival link in addition to the URL.
  2. Notation for 'data operators' is introduced without a compact tabular summary of all operators considered; adding such a table would improve readability when the framework is applied to new research questions.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback highlighting areas where the framework's presentation and evidential basis can be strengthened. We address each major comment below, proposing targeted revisions to improve clarity, structure, and operational guidance while preserving the paper's core contribution as a retrospective strategic guide.

read point-by-point responses

  1. Referee: [§4 and §5] §4 (framework distillation) and §5 (KITScenes case study): The central claim that the diagnosis step plus minimal-operator selection reliably yields more impactful datasets than alternatives rests on post-hoc reframing of existing datasets and a single self-authored case study. No forward test, controlled comparison against alternative design processes, or external replication is reported, so the optimality of the minimality criterion remains unverified.

    Authors: We agree that the framework's support is retrospective, drawn from historical dataset analysis and our KITScenes case study, without prospective or controlled validation of optimality. The manuscript presents this as a distilled strategic approach rather than an empirically proven optimal method. In revision we will add an explicit limitations subsection in §4 or §6 clarifying the evidential basis, tempering claims about reliability, and outlining the need for future forward tests or external replications. This addresses the concern without requiring new experiments. revision: partial

  2. Referee: [§3] §3 (evolution analysis): The mapping of historical dataset decisions onto the proposed 'data operator' taxonomy is presented as evidence for the framework, but the taxonomy itself is introduced in the same section; this creates a risk that the analysis is shaped by the framework rather than independently motivating it.

    Authors: The taxonomy was derived bottom-up from patterns observed across dataset histories, but we recognize the risk of circular presentation. We will restructure §3 to first describe the raw evolution and decision patterns in major AD datasets independently of the taxonomy, then introduce the taxonomy as a formalization of those patterns, and finally map the datasets onto it. This separation will make the independent motivation explicit. revision: yes

  3. Referee: [§2] §2 (gap identification): The distinction between 'data problem' and 'evaluation problem' is introduced without an operational, reproducible procedure or decision criteria; without such a procedure the diagnosis step cannot be applied consistently by other teams, undermining the claim that the framework is strategic and generalizable.

    Authors: We accept that greater operational detail is needed for reproducibility. In the revised §2 we will include explicit decision criteria, a step-by-step checklist, and a simple flowchart for distinguishing data versus evaluation problems, illustrated with examples drawn from the historical analysis in §3. This will make the diagnosis step actionable for other teams. revision: yes

Circularity Check

0 steps flagged

No significant circularity; framework derived from external dataset analysis

full rationale

The paper derives its strategic framework by analyzing the evolution of major existing autonomous driving datasets through the proposed diagnosis-and-minimal-operator lens, then illustrates the framework via its own KITScenes case study. No equations, fitted parameters, or self-citation chains are present that reduce any central claim to its own inputs by construction. The derivation remains self-contained against the analyzed external datasets and does not rely on renaming, smuggling ansatzes, or load-bearing self-references that would force the result.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The paper relies on domain assumptions about problem classification and introduces the concept of data operators as a new way to think about dataset design.

axioms (2)
  • domain assumption Research questions in autonomous driving can be classified as blocked by either a data problem or an evaluation problem.
    This classification is the starting point of the framework as stated in the abstract.
  • ad hoc to paper Selecting the minimal data operator(s) is the optimal way to close the gap.
    The framework assumes this is the way to proceed after diagnosis.
invented entities (1)
  • data operator no independent evidence
    purpose: A method or action to address data or evaluation gaps without necessarily collecting new data.
    Introduced as part of the framework to describe ways to close gaps.

pith-pipeline@v0.9.1-grok · 5783 in / 1507 out tokens · 35394 ms · 2026-07-02T14:13:18.836896+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

52 extracted references · 9 canonical work pages · 3 internal anchors

  1. [1]

    The Road Ahead in Autonomous Driving: The KITScenes Multimodal Dataset

    R. Schwarzkopf, F. Immel, A. Blumberg, J. Merkert, N. Rack, K. Wang, F. Konstantinidis, J. Truetsch, C. Fernandez, A. B¨ atz, K. R¨ osch, M. Steiner, W. Poh, Y. Shen, R. Wagner, F. Hauser, D. Strutz, J. Villa, G. Stepanov, H. Caesar, ¨Omer S ¸ahin Ta¸ s, F. Bieder, J.-H. Pauls, and C. Stiller, “The road ahead in autonomous driving: The kitscenes multimoda...

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  2. [2]

    LongTail Driving Scenarios with Reasoning Traces: The KITScenes LongTail Dataset

    R. Wagner, O. S. Tas, J. Villa, F. Hauser, Y. Shen, M. Steiner, D. Strutz, C. Fernandez, C. Kinzig, G. S. Guitierrez-Cabello, H. K¨ onigshof, F. Immel, R. Schwarzkopf, N. A. Rack, K. R¨ osch, K. Wang, J.-H. Pauls, M. Lauer, I. Gilitschenski, H. Caesar, and C. Stiller, “Longtail driving scenarios with reasoning traces: The kitscenes longtail dataset,” 2026...

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  3. [3]

    Argoverse 2: Next generation datasets for self- driving perception and forecasting,

    B. Wilson, W. Qi, T. Agarwal, J. Lambert, J. Singh, S. Khandelwal, B. Pan, R. Kumar, A. Hartnett, J. K. Pontes, D. Ramanan, P. Carr, and J. Hays, “Argoverse 2: Next generation datasets for self- driving perception and forecasting,” inProceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks (NeurIPS Datasets and Benchmarks ...

    2021

  4. [4]

    NVIDIA Autonomous Vehicle Dataset,

    PhysicalAI Autonomous Vehicles, “NVIDIA Autonomous Vehicle Dataset,” 2025, accessed: 2026-01-

    2025

  5. [5]

    Available: https://huggingface.co/datasets/nvidia/PhysicalAI-Autonomous-Vehicles

    [Online]. Available: https://huggingface.co/datasets/nvidia/PhysicalAI-Autonomous-Vehicles

  6. [6]

    Towards learning-based planning: The nuplan benchmark for real-world autonomous driving,

    N. Karnchanachari, D. Geromichalos, K. S. Tan, N. Li, C. Eriksen, S. Yaghoubi, N. Mehdipour, G. Bernasconi, W. K. Fong, Y. Guo, and H. Caesar, “Towards learning-based planning: The nuplan benchmark for real-world autonomous driving,” in2024 IEEE International Conference on Robotics and Automation (ICRA), 2024, pp. 629–636

    2024

  7. [7]

    Vision-based End-to-End Driving Challenge 2025,

    Waymo Open Dataset, “Vision-based End-to-End Driving Challenge 2025,” 2025, accessed: 2025-11-01. [Online]. Available: https://waymo.com/open/challenges/2025/e2e-driving

    2025

  8. [8]

    Are we ready for autonomous driving? the kitti vision bench- mark suite,

    A. Geiger, P. Lenz, and R. Urtasun, “Are we ready for autonomous driving? the kitti vision bench- mark suite,” in2012 IEEE Conference on Computer Vision and Pattern Recognition, 2012, pp. 3354–3361

    2012

  9. [9]

    Virtual worlds as proxy for multi-object tracking analysis,

    A. Gaidon, Q. Wang, Y. Cabon, and E. Vig, “Virtual worlds as proxy for multi-object tracking analysis,” inCVPR, 2016

    2016

  10. [10]

    nuscenes: A multimodal dataset for autonomous driving,

    H. Caesar, V. Bankiti, A. H. Lang, S. Vora, V. E. Liong, Q. Xu, A. Krishnan, Y. Pan, G. Baldan, and O. Beijbom, “nuscenes: A multimodal dataset for autonomous driving,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2020

    2020

  11. [11]

    Scalability in perception for autonomous driving: Waymo open dataset,

    P. Sun, H. Kretzschmar, X. Dotiwalla, A. Chouard, V. Patnaik, P. Tsui, J. Guo, Y. Zhou, Y. Chai, B. Caine, V. Vasudevan, W. Han, J. Ngiam, H. Zhao, A. Timofeev, S. Ettinger, M. Krivokon, A. Gao, A. Joshi, Y. Zhang, J. Shlens, Z. Chen, and D. Anguelov, “Scalability in perception for autonomous driving: Waymo open dataset,” inProceedings of the IEEE/CVF Con...

    2020

  12. [12]

    Large scale interactive motion forecasting for autonomous driving: The waymo open motion da- taset,

    S. Ettinger, S. Cheng, B. Caine, C. Liu, H. Zhao, S. Pradhan, Y. Chai, B. Sapp, C. R. Qi, Y. Zhou, Z. Yang, A. Chouard, P. Sun, J. Ngiam, V. Vasudevan, A. McCauley, J. Shlens, and D. Anguelov, “Large scale interactive motion forecasting for autonomous driving: The waymo open motion da- taset,” inProceedings of the IEEE/CVF International Conference on Comp...

    2021

  13. [13]

    BDD100K: A diverse driving dataset for heterogeneous multitask learning,

    F. Yu, H. Chen, X. Wang, W. Xian, Y. Chen, F. Liu, V. Madhavan, and T. Darrell, “BDD100K: A diverse driving dataset for heterogeneous multitask learning,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020

    2020

  14. [14]

    The mapillary vistas dataset for semantic understanding of street scenes,

    G. Neuhold, T. Ollmann, S. Rota Bul` o, and P. Kontschieder, “The mapillary vistas dataset for semantic understanding of street scenes,” inProceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 5000–5009

    2017

  15. [15]

    Openlane- v2: A topology reasoning benchmark for unified 3d hd mapping,

    H. Wang, T. Li, Y. Li, L. Chen, C. Sima, Z. Liu, B. Wang, P. Jia, Y. Wang, S. Jianget al., “Openlane- v2: A topology reasoning benchmark for unified 3d hd mapping,” inThirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track, 2023

    2023

  16. [16]

    Argotweak: Towards self-updating hd maps through struc- tured priors,

    L. Wild, R. Valencia, and P. Jensfelt, “Argotweak: Towards self-updating hd maps through struc- tured priors,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2025

    2025

  17. [17]

    123D: Unifying Multi-Modal Autonomous Driving Data at Scale

    D. Dauner, V. Charraut, B. Berle, T. Li, L. Nguyen, J. Wang, C. Jing, M. Igl, H. Caesar, B. Ivanovic, A. Geiger, and K. Chitta, “123d: Unifying multi-modal autonomous driving data at scale,”arXiv preprint arXiv:2605.08084, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  18. [18]

    Cosmos-drive-dreams: Scalable synthetic driving data generation with world foundation models,

    X. Ren, Y. Lu, T. Cao, R. Gao, S. Huang, A. Sabour, T. Shen, T. Pfaff, J. Z. Wu, R. Chen, S. W. Kim, J. Gao, L. Leal-Taixe, M. Chen, S. Fidler, and H. Ling, “Cosmos-drive-dreams: Scalable synthetic driving data generation with world foundation models,” 2025. [Online]. Available: https://arxiv.org/abs/2506.09042

    work page arXiv 2025

  19. [19]

    Lanelet2: A high-definition map framework for the future of automated driving,

    F. Poggenhans, J.-H. Pauls, J. Janosovits, S. Orf, M. Naumann, F. Kuhnt, and M. Mayr, “Lanelet2: A high-definition map framework for the future of automated driving,” in2018 21st International Conference on Intelligent Transportation Systems (ITSC), Hawaii, USA, November 2018, pp. 1672–1679. [Online]. Available: http://www.mrt.kit.edu/z/publ/download/ 201...

    2018

  20. [20]

    The cityscapes dataset for semantic urban scene understanding,

    M. Cordts, M. Omran, S. Ramos, T. Rehfeld, M. Enzweiler, R. Benenson, U. Franke, S. Roth, and B. Schiele, “The cityscapes dataset for semantic urban scene understanding,” inProc. of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016

    2016

  21. [21]

    KITTI-360: A novel dataset and benchmarks for urban scene understanding in 2d and 3d,

    Y. Liao, J. Xie, and A. Geiger, “KITTI-360: A novel dataset and benchmarks for urban scene understanding in 2d and 3d,”Pattern Analysis and Machine Intelligence (PAMI), 2022

    2022

  22. [22]

    Se- manticKITTI: A dataset for semantic scene understanding of LiDAR sequences,

    J. Behley, M. Garbade, A. Milioto, J. Quenzel, S. Behnke, C. Stachniss, and J. Gall, “Se- manticKITTI: A dataset for semantic scene understanding of LiDAR sequences,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2019

    2019

  23. [23]

    The ApolloScape open dataset for autonomous driving and its application,

    X. Huang, P. Wang, X. Cheng, D. Zhou, Q. Geng, and R. Yang, “ The ApolloScape Open Dataset for Autonomous Driving and Its Application ,”IEEE Transactions on Pattern Analysis & Machine Intelligence, vol. 42, no. 10, pp. 2702–2719, Oct. 2020. [Online]. Available: https://doi.ieeecomputersociety.org/10.1109/TPAMI.2019.2926463

    work page doi:10.1109/tpami.2019.2926463 2020

  24. [24]

    CARLA: An open urban driving simulator,

    A. Dosovitskiy, G. Ros, F. Codevilla, A. Lopez, and V. Koltun, “CARLA: An open urban driving simulator,” inProceedings of the 1st Annual Conference on Robot Learning (CoRL), ser. Proceedings of Machine Learning Research, vol. 78. PMLR, 2017, pp. 1–16

    2017

  25. [25]

    Bench2drive: Towards multi-ability benchmarking of closed-loop end-to-end autonomous driving,

    X. Jia, Z. Yang, Q. Li, Z. Zhang, and J. Yan, “Bench2drive: Towards multi-ability benchmarking of closed-loop end-to-end autonomous driving,” inNeurIPS 2024 Datasets and Benchmarks Track, 2024

    2024

  26. [26]

    Argoverse: 3d tracking and forecasting with rich maps,

    M.-F. Chang, J. W. Lambert, P. Sangkloy, J. Singh, S. Bak, A. Hartnett, D. Wang, P. Carr, S. Lucey, D. Ramanan, and J. Hays, “Argoverse: 3d tracking and forecasting with rich maps,” inConference on Computer Vision and Pattern Recognition (CVPR), 2019

    2019

  27. [27]

    One thousand and one hours: Self-driving motion prediction dataset,

    J. Houston, G. Zuidhof, L. Bergamini, Y. Ye, L. Chen, A. Jain, S. Omari, V. Iglovikov, and P. On- druska, “One thousand and one hours: Self-driving motion prediction dataset,” inProceedings of the 2020 Conference on Robot Learning (CoRL), ser. Proceedings of Machine Learning Research, vol

    2020

  28. [28]

    PMLR, 2021, pp. 409–418

    2021

  29. [29]

    One million scenes for autonomous driving: Once dataset,

    J. Mao, N. Minzhe, C. Jiang, h. liang, J. Chen, X. Liang, Y. Li, C. Ye, W. Zhang, Z. Li, J. Yu, C. XU, and H. Xu, “One million scenes for autonomous driving: Once dataset,” inProceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks, J. Vanschoren and S. Yeung, Eds., vol. 1, 2021. [Online]. Available: https://datasets-bench...

    2021

  30. [30]

    Zenseact open dataset: A large-scale and diverse multimodal dataset for autonomous driving,

    M. Alibeigi, W. Ljungbergh, A. Tonderski, G. Hess, A. Lilja, C. Lindstrom, D. Motorniuk, J. Fu, J. Widahl, and C. Petersson, “Zenseact open dataset: A large-scale and diverse multimodal dataset for autonomous driving,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023

    2023

  31. [31]

    PandaSet: Advanced sensor suite dataset for autonomous driving,

    P. Xiao, Z. Shao, S. Hao, Z. Zhang, X. Chai, J. Jiao, Z. Li, J. Wu, K. Sun, K. Jiang, Y. Wang, and D. Yang, “PandaSet: Advanced sensor suite dataset for autonomous driving,” in2021 IEEE International Intelligent Transportation Systems Conference (ITSC), 2021, pp. 3095–3101

    2021

  32. [32]

    A2D2: Audi autonomous driving dataset,

    J. Geyer, Y. Kassahun, M. Mahmudi, X. Ricou, R. Durgesh, A. S. Chung, L. Hauswald, V. H. Pham, M. M¨ uhlegg, S. Dorn, T. Fernandez, M. J¨ anicke, S. Mirashi, C. Savani, M. Sturm, O. Vorobiov, M. Oelker, S. Garreis, and P. Schuberth, “A2D2: Audi autonomous driving dataset,”

  33. [33]

    Available: https://www.a2d2.audi

    [Online]. Available: https://www.a2d2.audi

  34. [34]

    A commute in data: The comma2k19 dataset,

    H. Schafer, E. Santana, A. Haden, and R. Biasini, “A commute in data: The comma2k19 dataset,” 2018

    2018

  35. [35]

    NAVSIM: Data-driven non-reactive autonomous vehicle simulation and benchmarking,

    D. Dauner, M. Hallgarten, T. Li, X. Weng, Z. Huang, Z. Yang, H. Li, I. Gilitschenski, B. Ivanovic, M. Pavone, A. Geiger, and K. Chitta, “NAVSIM: Data-driven non-reactive autonomous vehicle simulation and benchmarking,” inAdvances in Neural Information Processing Systems (NeurIPS), 2024

    2024

  36. [36]

    NeuroNCAP: Photorealistic closed-loop safety testing for autonomous driving,

    W. Ljungbergh, A. Tonderski, J. Johnander, H. Caesar, K. ˚Astr¨ om, M. Felsberg, and C. Petersson, “NeuroNCAP: Photorealistic closed-loop safety testing for autonomous driving,” inProceedings of the European Conference on Computer Vision (ECCV). Springer, 2024, pp. 161–177

    2024

  37. [37]

    Maptr: Structured modeling and learning for online vectorized hd map construction,

    B. Liao, S. Chen, X. Wang, T. Cheng, Q. Zhang, W. Liu, and C. Huang, “Maptr: Structured modeling and learning for online vectorized hd map construction,” inThe Eleventh International Conference on Learning Representations, 2022

    2022

  38. [38]

    Maptrv2: An end-to-end framework for online vectorized hd map construction,

    B. Liao, S. Chen, Y. Zhang, B. Jiang, Q. Zhang, W. Liu, C. Huang, and X. Wang, “Maptrv2: An end-to-end framework for online vectorized hd map construction,”International Journal of Computer Vision, Oct 2024. [Online]. Available: https://doi.org/10.1007/s11263-024-02235-z

    work page doi:10.1007/s11263-024-02235-z 2024

  39. [39]

    Streammapnet: Streaming mapping network for vectorized online hd map construction,

    T. Yuan, Y. Liu, Y. Wang, Y. Wang, and H. Zhao, “Streammapnet: Streaming mapping network for vectorized online hd map construction,” inProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2024, pp. 7356–7365

    2024

  40. [40]

    Stream query denoising for vectorized hd-map construction,

    S. Wang, F. Jia, W. Mao, Y. Liu, Y. Zhao, Z. Chen, T. Wang, C. Zhang, X. Zhang, and F. Zhao, “Stream query denoising for vectorized hd-map construction,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 203–220

    2024

  41. [41]

    Maptracker: Tracking with strided memory fusion for consistent vector hd mapping,

    J. Chen, Y. Wu, J. Tan, H. Ma, and Y. Furukawa, “Maptracker: Tracking with strided memory fusion for consistent vector hd mapping,” inComputer Vision – ECCV 2024, A. Leonardis, E. Ricci, S. Roth, O. Russakovsky, T. Sattler, and G. Varol, Eds. Cham: Springer Nature Switzerland, 2025, pp. 90–107

    2024

  42. [42]

    Enhancing vectorized map perception with historical rasterized maps,

    X. Zhang, G. Liu, Z. Liu, N. Xu, Y. Liu, and J. Zhao, “Enhancing vectorized map perception with historical rasterized maps,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 422–439

    2024

  43. [43]

    Globalmapnet: An online framework for vectorized global hd map construction,

    A. Shi, Y. Cai, X. Chen, J. Pu, Z. Fu, and H. Lu, “Globalmapnet: An online framework for vectorized global hd map construction,”arXiv preprint arXiv:2409.10063, 2024

    work page arXiv 2024

  44. [44]

    Mapexpert: Online hd map construction with simple and efficient sparse map element expert,

    D. Zhang, D. Chen, P. Zhi, Y. Chen, Z. Yuan, C. Li, R. Zhou, Q. Zhouet al., “Mapexpert: Online hd map construction with simple and efficient sparse map element expert,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 14, 2025, pp. 14 745–14 753

    2025

  45. [45]

    Histrackmap: Global vectorized high-definition map construction via history map tracking,

    J. Yang, S. Yang, X. Tan, and H. Wang, “Histrackmap: Global vectorized high-definition map construction via history map tracking,”arXiv preprint arXiv:2503.07168, 2025

    work page arXiv 2025

  46. [46]

    SDTagnet: Leveraging text-annotated navigation maps for online HD map construction,

    F. Immel, J.-H. Pauls, R. Fehler, F. Bieder, J. Merkert, and C. Stiller, “SDTagnet: Leveraging text-annotated navigation maps for online HD map construction,” inThe Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025. [Online]. Available: https://openreview.net/forum?id=N3E1cU8Cv3

    2025

  47. [47]

    Mapping like a skeptic: Probabilistic bev projection for online hd mapping,

    F. Erdo˘ gan, M. R. Barın, and F. G¨ uney, “Mapping like a skeptic: Probabilistic bev projection for online hd mapping,”arXiv preprint arXiv:2508.21689, 2025

    work page arXiv 2025

  48. [48]

    Gtsign-220: A crowd-sourced, stvo-aligned benchmark for fine-grained german traffic sign recognition,

    M. L. Carnot, E. Fastermann, J. Kunze, E. Peukert, A. Ludwig, and B. Franczyk, “Gtsign-220: A crowd-sourced, stvo-aligned benchmark for fine-grained german traffic sign recognition,” inIntelli- gent Vehicles Symposium (IV), 2026

    2026

  49. [49]

    Automatic mapping of tailored landmark representations for automated driving and map learning,

    J.-H. Pauls, B. Schmidt, and C. Stiller, “Automatic mapping of tailored landmark representations for automated driving and map learning,” in2021 IEEE International Conference on Robotics and Automation (ICRA), 2021, pp. 6725–6731

    2021

  50. [50]

    Autoware,

    Autoware Foundation, “Autoware,” https://github.com/autowarefoundation/autoware, accessed: 2026-05-02

    2026

  51. [51]

    ASAM e.V.,ASAM OpenDRIVE 1.8.0 Specification, November 2023, published November 22,

    2023

  52. [52]

    Available: https://www.asam.net/standards/detail/opendrive/

    [Online]. Available: https://www.asam.net/standards/detail/opendrive/